Enabling pipeline configuration in DU

---

Document Understanding (DU) platform now provides you the flexibility to choose your pipeline during project set up. This helps you to optimize the hardware requirements depending on the customer needs and thereby offer competitive prices.

DU includes two active pipelines VESPA & SmartExtract and both of them are capable of extracting Insights & Table fields of interest from the documents. However to get maximum advantage the recommended usage is as below.

1. Customers dealing with Forms type of documents which contains similar information, structured/semi-structured data, labeled values, tables should use SmartExtract pipeline. (Document type examples - Invoices, Purchase Order)
2. Customers requiring long text extraction from documents with paragraphs & sections should use VESPA pipeline. (Document type examples – Contract/Legal documents)
3. Customers dealing with both types of documents should use a combination of VESPA & SmartExtract pipelines.

Support customization of VESPA pre-processors

---

We introduced the VESPA Field configurator feature in 7.1 release which enables user to add new fields/modify existing fields depending on the customer needs. We have improved the flexibility of this feature by including the provision for custom pre-processors. Pre-processors helps to fine tune the extraction process and thereby yields better accuracy. This feature allows you to view the available pre-processors and modify/add new ones to suite the specific requirements of the customer. This is applicable for VESPA pipeline.

Pre-trained models

---

EITL model trained with Navistar production invoices

Today, with improved DU capabilities we are catering to document types beyond invoices and there is a compelling need to have a generic PWF that can support multiple types of documents. So we introduce, SmartVision PWF which offers you the basic review & approve capability for any document type. In fact, it is a generalized version of Invoice Extraction PWF and includes all the functionalities which is currently available in IE PWF.

Personas – SVision Supervisor, SVision Reviewer, Installation Engineer

Document Types – Any (as set in the attached DU project)

Functionalities

---

• Ability to set up multiple PWF projects to support different document types
• Document listing dashboard that provides the operational summary view, multiple search options & time zone based display
• Automated document allocation to specific list of users
• Document review & correction
• Exception processing workflows – Reject/Escalate/New Invoices/Delete
• Document export based on pre-defined schedule
• Advanced Reporting & Analytics

Additionally non-functional requirements supported today will continue with SmartVision PWF

Feature Enhancements

---

• Provision to configure the users who needs to participate in auto allocation process. This gives the flexibility to add/remove users from auto allocation process depending on the business needs
• Provision to persist custom layer data in PWF and use in conjunction with system extracted data
• Review screen enhancements to improve productivity of reviewer
  • Systematically identify duplicate invoices and notify the reviewer during the review process. This helps the reviewer to timely reject the invoices
  • Extend Point & Crop feature for newly added/split invoices also
  • Re-order the display sequence of Insights & Line Items so as to help the reviewer to focus on the most commonly entered fields